Llama 3.1 405B.
Meta's largest open-weight LLM — dense 405B, frontier-class at launch.
1× AMD MI325.
Most-aggressive quantisation we have a working recommendation for. Lower precision = less VRAM = cheaper hardware, at a small accuracy cost.
Cheapest hosted endpoints.
| Provider | Access | $/M in | $/M out | |
|---|---|---|---|---|
| Self-hosted on rented GPU cluster | self hosted | — | — | Run yourself → |
| Together AI | hosted inference | — | — | Launch ↗ |
| Ollama | self hosted | — | — | Launch ↗ |
What it's best at.
Variants in the Llama family.
Llama 3.3 70B
Meta's best-in-class open-weight LLM — 70B class.
Llama 3.1 70B
Llama 3.1 70B — production workhorse, superseded by 3.3 but still widely depl...
Llama 3.1 8B
Meta's most popular open-weight small LLM — fits anywhere.
Llama 3.2 90B Vision
Meta's largest vision-capable Llama.
Llama 3.2 11B Vision
Meta's open-weight multimodal LLM — vision + text in 11B.
Llama 3.2 3B
3B Llama — laptop-class chat + RAG.
Llama 3.2 1B
Meta's smallest Llama — mobile + on-device target.
Workloads.
Frequently asked.
How do I run Llama 3.1 405B?
Where can I access Llama 3.1 405B?
How much does it cost to run Llama 3.1 405B?
Is Llama 3.1 405B open-source or proprietary?
Cheapest hardware per quantisation.
Each row is one quantisation tier (the same weights compressed differently). Lower precision → lower VRAM → cheaper hardware, at the cost of small accuracy loss. $/hr refreshed hourly from each provider's API.
| Quantisation | Cheapest GPU config | Total VRAM | Live $/hr | tokens/sec | |
|---|---|---|---|---|---|
|
FP16
FP16 — half precision (default)
|
1024 GB | — | — | Compare → | |
|
FP8
FP8 — 8-bit float (Hopper / Blackwell)
|
512 GB | — | — | Compare → | |
|
INT8
INT8 — 8-bit integer
|
640 GB | $3.84/hr | — | Compare → | |
|
INT4
INT4 — 4-bit integer (~4× VRAM saving)
|
256 GB | — | — | Compare → |
What it costs per month across providers.
Estimate your monthly bill for Llama 3.1 405B across every host that publishes per-token pricing. Slide your token volumes; the chart + table re-rank cheapest-first.
No priced API access rows on file for Llama 3.1 405B yet.
Rent the GPU instead of paying per token.
For an open-weights model like Llama 3.1 405B, you can rent a GPU and serve inference yourself. The math: cheapest GPU rental × 730 hours/month + your electricity rate × power draw.
Assumes the GPU runs 24/7 at ~85% utilisation. If your traffic is bursty, you'll pay less for the API and probably more for the GPU (idle hours still cost rental). The breakeven analysis lives on the Self-host vs API breakeven tool.
What it's best at.
Scores normalised against benchmark ceilings (100 = perfect). Coloured by tier — coral 80+ frontier, lavender 65+ strong, sage 50+ solid, slate below.
Published scores.
| Benchmark | Score | Source |
|---|---|---|
| GPQA | 51.1 | official ↗ |
| MATH | 73.8 | official ↗ |
| MMLU | 88.6 | official ↗ |
| IFEval | 88.6 | official ↗ |
| MMLU-Pro | 73.3 | official ↗ |
| HumanEval | 89.0 | official ↗ |
Independent rankings.
About Llama 3.1 405B.
Llama 3.1 405B is Meta's largest open-weight model — dense (not MoE), 405B parameters trained on 15T tokens with 30M H100-hours of compute. At launch (July 2024) it was the first open model competitive with GPT-4 and Claude 3.5 Sonnet on most reasoning benchmarks. Mostly superseded by Llama 3.3 70B for production (same quality at 1/6th the inference cost), but still relevant for research and as a teacher model for distillation. Runs on 8× H100 (FP16) or 4× H100 (FP8).